Nightmare is a high-level browser automation library from Segment.

The goal is to expose a few simple methods that mimic user actions (like goto, type and click), with an API that feels synchronous for each block of scripting, rather than deeply nested callbacks. It was originally designed for automating tasks across sites that don't have APIs, but is most often used for UI testing and crawling.

Under the covers it uses Electron, which is similar to PhantomJS but roughly twice as fast and more modern. Because Nightmare uses Electron, it is your responsibility to ensure that the webpages loaded by Nightmare are not malicious. If you do load a malicious website, that website can execute arbitrary code on your computer.

Niffy is a perceptual diffing tool built on Nightmare. It helps you detect UI changes and bugs across releases of your web app.

Daydream is a complementary chrome extension built by @stevenmiller888 that generates Nightmare scripts for you while you browse.

Many thanks to @matthewmueller and @rosshinkley for their help on Nightmare.

• Examples
  • UI Testing Quick Start
  • Perceptual Diffing with Niffy & Nightmare
• API
  • Set up an instance
  • Interact with the page
  • Extract from the page
  • Cookies
  • Proxies
  • Promises
  • Extending Nightmare
• Usage
• Debugging
• Supplementary Examples and Documentation

Let's search on DuckDuckGo:

import Nightmare from 'nightmare';
const nightmare = Nightmare({ show: true });

nightmare
  .goto('https://duckduckgo.com')
  .type('#search_form_input_homepage', 'github nightmare')
  .click('#search_button_homepage')
  .wait('#r1-0 a.result__a')
  .evaluate(() => document.querySelector('#r1-0 a.result__a').href)
  .end()
  .then(console.log)
  .catch((error) => {
    console.error('Search failed:', error);
  });

You can run this with:

npm install --save nightmare
node example.js

Or, let's run some mocha tests:

import Nightmare from 'nightmare';
import {expect} from 'chai';

describe('test duckduckgo search results', () => {
  it('should find the nightmare github link first', (done) => {
    const nightmare = Nightmare()
    nightmare
      .goto('https://duckduckgo.com')
      .type('#search_form_input_homepage', 'github nightmare')
      .click('#search_button_homepage')
      .wait('#zero_click_wrapper .c-info__title a')
      .evaluate(() =>
        document.querySelector('#zero_click_wrapper .c-info__title a').href
      )
      .end()
      .then((link) => {
        expect(link).to.equal('https://github.com/segmentio/nightmare');
        done();
      })
  });
});

You can see examples of every function in the tests here.

Please note that the examples are using the mocha-generators package for Mocha, which enables the support for generators.

To get started with UI Testing, check out this quick start guide.

npm install

npm test

Nightmare is intended to be run on NodeJS 4.x or higher.

Creates a new instance that can navigate around the web. The available options are documented here, along with the following nightmare-specific options.

Returns the version of Nightmare.

Throws an exception if the .wait() didn't return true within the set timeframe.

const nightmare = Nightmare({
  waitTimeout: 1000 // in ms
});

Throws an exception if the .goto() didn't finish loading within the set timeframe. Note that, even though goto normally waits for all the resources on a page to load, a timeout exception is only raised if the DOM itself has not yet loaded.

const nightmare = Nightmare({
  gotoTimeout: 1000 // in ms
});

Forces Nightmare to move on if a page transition caused by an action (eg, .click()) didn't finish within the set timeframe. If loadTimeout is shorter than gotoTimeout, the exceptions thrown by gotoTimeout will be suppressed.

const nightmare = Nightmare({
  loadTimeout: 1000 // in ms
});

The maximum amount of time to wait for an .evaluate() statement to complete.

const nightmare = Nightmare({
  executionTimeout: 1000 // in ms
});

The default system paths that Electron knows about. Here's a list of available paths: https://github.com/atom/electron/blob/master/docs/api/app.md#appgetpathname

You can overwrite them in Nightmare by doing the following:

const nightmare = Nightmare({
  paths: {
    userData: '/user/data'
  }
});

The command line switches used by the Chrome browser that are also supported by Electron. Here's a list of supported Chrome command line switches: https://github.com/atom/electron/blob/master/docs/api/chrome-command-line-switches.md

const nightmare = Nightmare({
  switches: {
    'proxy-server': '1.2.3.4:5678',
    'ignore-certificate-errors': true
  }
});

The path to the prebuilt Electron binary. This is useful for testing on different versions of Electron. Note that Nightmare only supports the version on which this package depends. Use this option at your own risk.

const nightmare = Nightmare({
  electronPath: require('electron')
});

A boolean to optionally show the Electron icon in the dock (defaults to false). This is useful for testing purposes.

const nightmare = Nightmare({
  dock: true
});

Optionally shows the DevTools in the Electron window using true, or use an object hash containing mode: 'detach' to show in a separate window. The hash gets passed to contents.openDevTools() to be handled. This is also useful for testing purposes. Note that this option is honored only if show is set to true.

const nightmare = Nightmare({
  openDevTools: {
    mode: 'detach'
  },
  show: true
});

How long to wait between keystrokes when using .type().

const nightmare = Nightmare({
  typeInterval: 20
});

How long to wait between checks for the .wait() condition to be successful.

const nightmare = Nightmare({
  pollInterval: 50 //in ms
});

Defines the number of times to retry an authentication when set up with .authenticate().

const nightmare = Nightmare({
  maxAuthRetries: 3
});

Gets the versions for Electron and Chromium.

Sets the useragent used by electron.

Sets the user and password for accessing a web page using basic authentication. Be sure to set it before calling .goto(url).

Completes any queue operations, disconnect and close the electron process. Note that if you're using promises, .then() must be called after .end() to run the .end() task. Also note that if using an .end() callback, the .end() call is equivalent to calling .end() followed by .then(fn). Consider:

nightmare
  .goto(someUrl)
  .end(() => "some value")
  //prints "some value"
  .then(console.log);

Clears all queued operations, kills the electron process, and passes error message or 'Nightmare Halted' to an unresolved promise. Done will be called after the process has exited.

Loads the page at url. Optionally, a headers hash can be supplied to set headers on the goto request.

When a page load is successful, goto returns an object with metadata about the page load, including:

• url: The URL that was loaded
• code: The HTTP status code (e.g. 200, 404, 500)
• method: The HTTP method used (e.g. "GET", "POST")
• referrer: The page that the window was displaying prior to this load or an empty string if this is the first page load.
• headers: An object representing the response headers for the request as in {header1-name: header1-value, header2-name: header2-value}

If the page load fails, the error will be an object with the following properties:

• message: A string describing the type of error
• code: The underlying error code describing what went wrong. Note this is NOT the HTTP status code. For possible values, see https://code.google.com/p/chromium/codesearch#chromium/src/net/base/net_error_list.h
• details: A string with additional details about the error. This may be null or an empty string.
• url: The URL that failed to load

Note that any valid response from a server is considered “successful.” That means things like 404 “not found” errors are successful results for goto. Only things that would cause no page to appear in the browser window, such as no server responding at the given address, the server hanging up in the middle of a response, or invalid URLs, are errors.

You can also adjust how long goto will wait before timing out by setting the gotoTimeout option on the Nightmare constructor.

Goes back to the previous page.

Goes forward to the next page.

Refreshes the current page.

Clicks the selector element once.

Mousedowns the selector element once.

Mouseups the selector element once.

Mouseovers the selector element once.

Mouseout the selector element once.

Enters the text provided into the selector element. Empty or falsey values provided for text will clear the selector's value.

.type() mimics a user typing in a textbox and will emit the proper keyboard events.

Key presses can also be fired using Unicode values with .type(). For example, if you wanted to fire an enter key press, you would write .type('body', '\u000d').

If you don't need the keyboard events, consider using .insert() instead as it will be faster and more robust.

Similar to .type(), .insert() enters the text provided into the selector element. Empty or falsey values provided for text will clear the selector's value.

.insert() is faster than .type() but does not trigger the keyboard events.

Checks the selector checkbox element.

Unchecks the selector checkbox element.

Changes the selector dropdown element to the option with attribute [value=option]

Scrolls the page to desired position. top and left are always relative to the top left corner of the document.

Sets the viewport size.

Injects a local file onto the current page. The file type must be either js or css.

Invokes fn on the page with arg1, arg2,.... All the args are optional. On completion it returns the return value of fn. Useful for extracting information from the page. Here's an example:

const selector = 'h1';
nightmare
  .evaluate((selector) => {
    // now we're executing inside the browser scope.
    return document.querySelector(selector).innerText;
   }, selector) // <-- that's how you pass parameters from Node scope to browser scope
  .then((text) => {
    // ...
  })

Error-first callbacks are supported as a part of evaluate(). If the arguments passed are one fewer than the arguments expected for the evaluated function, the evaluation will be passed a callback as the last parameter to the function. For example:

const selector = 'h1';
nightmare
  .evaluate((selector, done) => {
    // now we're executing inside the browser scope.
    setTimeout(() => done(null, document.querySelector(selector).innerText), 2000);
   }, selector)
  .then((text) => {
    // ...
  })

Note that callbacks support only one value argument (eg function(err, value)). Ultimately, the callback will get wrapped in a native Promise and only be able to resolve a single value.

Promises are also supported as a part of evaluate(). If the return value of the function has a then member, .evaluate() assumes it is waiting for a promise. For example:

const selector = 'h1';
nightmare
  .evaluate((selector) => (
    new Promise((resolve, reject) => {
      setTimeout(() => resolve(document.querySelector(selector).innerText), 2000);
    )}, selector)
  )
  .then((text) => {
    // ...
  })

Waits for ms milliseconds e.g. .wait(5000).

Waits until the element selector is present e.g. .wait('#pay-button').

Waits until the fn evaluated on the page with arg1, arg2,... returns true. All the args are optional. See .evaluate() for usage.

Adds a header override for all HTTP requests. If header is undefined, the header overrides will be reset.

Returns whether the selector exists or not on the page.

Returns whether the selector is visible or not.

Captures page events with the callback. You have to call .on() before calling .goto(). Supported events are documented here.

This event is triggered if any javascript exception is thrown on the page. But this event is not triggered if the injected javascript code (e.g. via .evaluate()) is throwing an exception.

Listens for window.addEventListener('error'), alert(...), prompt(...) & confirm(...).

Listens for top-level page errors. This will get triggered when an error is thrown on the page.

Nightmare disables window.alert from popping up by default, but you can still listen for the contents of the alert dialog.

Nightmare disables window.prompt from popping up by default, but you can still listen for the message to come up. If you need to handle the confirmation differently, you'll need to use your own preload script.

Nightmare disables window.confirm from popping up by default, but you can still listen for the message to come up. If you need to handle the confirmation differently, you'll need to use your own preload script.

type will be either log, warn or error and arguments are what gets passed from the console. This event is not triggered if the injected javascript code (e.g. via .evaluate()) is using console.log.

Similar to .on(), but captures page events with the callback one time.

Removes a given listener callback for an event.

Takes a screenshot of the current page. Useful for debugging. The output is always a png. Both arguments are optional. If path is provided, it saves the image to the disk. Otherwise it returns a Buffer of the image data. If clip is provided (as documented here), the image will be clipped to the rectangle.

Saves the current page as html as files to disk at the given path. Save type options are here.

Saves a PDF to the specified path. Options are here.

Returns the title of the current page.

Returns the url of the current page.

Returns the path name of the current page.

Gets a cookie by it's name. The url will be the current url.

Queries multiple cookies with the query object. If a query.name is set, it will return the first cookie it finds with that name, otherwise it will query for an array of cookies. If no query.url is set, it will use the current url. Here's an example:

// get all google cookies that are secure
// and have the path `/query`
nightmare
  .goto('http://google.com')
  .cookies.get({
    path: '/query',
    secure: true
  })
  .then((cookies) => {
    // do something with the cookies
  })

Available properties are documented here: https://github.com/atom/electron/blob/master/docs/api/session.md#sescookiesgetdetails-callback

Gets all the cookies for the current url. If you'd like get all cookies for all urls, use: .get({ url: null }).

Sets a cookie's name and value. This is the most basic form, and the url will be the current url.

Sets a cookie. If cookie.url is not set, it will set the cookie on the current url. Here's an example:

nightmare
  .goto('http://google.com')
  .cookies.set({
    name: 'token',
    value: 'some token',
    path: '/query',
    secure: true
  })
  // ... other actions ...
  .then(() => {
    // ...
  })

Available properties are documented here: https://github.com/atom/electron/blob/master/docs/api/session.md#sescookiessetdetails-callback

Sets multiple cookies at once. cookies is an array of cookie objects. Take a look at the .cookies.set(cookie) documentation above for a better idea of what cookie should look like.

Clears a cookie for the current domain. If name is not specified, all cookies for the current domain will be cleared.

nightmare
  .goto('http://google.com')
  .cookies.clear('SomeCookieName')
  // ... other actions ...
  .then(() => {
    // ...
  })

Clears all cookies for all domains.

nightmare
  .goto('http://google.com')
  .cookies.clearAll()
  // ... other actions ...
  .then(() => {
    //...
  });

Proxies are supported in Nightmare through switches.

If your proxy requires authentication you also need the authentication call.

The following example not only demonstrates how to use proxies, but you can run it to test if your proxy connection is working:

import Nightmare from 'nightmare';

const proxyNightmare = Nightmare({
  switches: {
    'proxy-server': 'my_proxy_server.example.com:8080' // set the proxy server here ...
  },
  show: true
});

proxyNightmare
  .authentication('proxyUsername', 'proxyPassword') // ... and authenticate here before `goto`
  .goto('http://www.ipchicken.com')
  .evaluate(() => {
    return document.querySelector('b').innerText.replace(/[^\d\.]/g, '');
  })
  .end()
  .then((ip) => { // This will log the Proxy's IP
    console.log('proxy IP:', ip);
  });

// The rest is just normal Nightmare to get your local IP
const regularNightmare = Nightmare({ show: true });

regularNightmare
  .goto('http://www.ipchicken.com')
  .evaluate(() =>
    document.querySelector('b').innerText.replace(/[^\d\.]/g, '');
  )
  .end()
  .then((ip) => { // This will log the your local IP
    console.log('local IP:', ip);
  });

By default, Nightmare uses default native ES6 promises. You can plug in your favorite ES6-style promises library like bluebird or q for convenience!

Here's an example:

var Nightmare = require('nightmare');

Nightmare.Promise = require('bluebird');
// OR:
Nightmare.Promise = require('q').Promise;

You can also specify a custom Promise library per-instance with the Promise constructor option like so:

var Nightmare = require('nightmare');

var es6Nightmare = Nightmare();
var bluebirdNightmare = Nightmare({
    Promise: require('bluebird')
});

var es6Promise = es6Nightmare.goto('https://github.com/segmentio/nightmare').then();
var bluebirdPromise = bluebirdNightmare.goto('https://github.com/segmentio/nightmare').then();

es6Promise.isFulfilled() // throws: `TypeError: es6EndPromise.isFulfilled is not a function`
bluebirdPromise.isFulfilled() // returns: `true | false`

You can add your own custom actions to the Nightmare prototype. Here's an example:

Nightmare.action('size', function(done) {
  this.evaluate_now(() => {
    const w = Math.max(document.documentElement.clientWidth, window.innerWidth || 0)
    const h = Math.max(document.documentElement.clientHeight, window.innerHeight || 0)
    return {
      height: h,
      width: w
    }
  }, done)
})

Nightmare()
  .goto('http://cnn.com')
  .size()
  .then((size) => {
    //... do something with the size information
  });

Remember, this is attached to the static class Nightmare, not the instance.

You'll notice we used an internal function evaluate_now. This function is different than nightmare.evaluate because it runs it immediately, whereas nightmare.evaluate is queued.

An easy way to remember: when in doubt, use evaluate. If you're creating custom actions, use evaluate_now. The technical reason is that since our action has already been queued and we're running it now, we shouldn't re-queue the evaluate function.

We can also create custom namespaces. We do this internally for nightmare.cookies.get and nightmare.cookies.set. These are useful if you have a bundle of actions you want to expose, but it will clutter up the main nightmare object. Here's an example of that:

Nightmare.action('style', {
  background(done) {
    this.evaluate_now(
      () => window.getComputedStyle(document.body, null).backgroundColor, done
    )
  }
})

Nightmare()
  .goto('http://google.com')
  .style.background()
  .then((background) => {
    // ... do something interesting with background  
  })

You can also add custom Electron actions. The additional Electron action or namespace actions take name, options, parent, win, renderer, and done. Note the Electron action comes first, mirroring how .evaluate() works. For example:

Nightmare.action('clearCache', (name, options, parent, win, renderer, done) => {
    parent.respondTo('clearCache', (done) => {
      win.webContents.session.clearCache(done);
    });
    done();
  },
  function (done) {
    this.child.call('clearCache', done);
  });

Nightmare()
  .clearCache()
  .goto('http://example.org')
  //... more actions ...
  .then(() => {
    // ...
  });

...would clear the browser’s cache before navigating to example.org.

See this document for more details on creating custom actions.

nightmare.use is useful for reusing a set of tasks on an instance. Check out nightmare-swiftly for some examples.

If you need to do something custom when you first load the window environment, you can specify a custom preload script. Here's how you do that:

import path from 'path';

const nightmare = Nightmare({
  webPreferences: {
    preload: path.resolve("custom-script.js")
    //alternative: preload: "absolute/path/to/custom-script.js"
  }
})

The only requirement for that script is that you'll need the following prelude:

window.__nightmare = {};
__nightmare.ipc = require('electron').ipcRenderer;

To benefit of all of nightmare's feedback from the browser, you can instead copy the contents of nightmare's preload script.

By default nightmare will create an in-memory partition for each instance. This means that any localStorage or cookies or any other form of persistent state will be destroyed when nightmare is ended. If you would like to persist state between instances you can use the webPreferences.partition api in electron.

import Nightmare from 'nightmare';

nightmare = Nightmare(); // non persistent paritition by default
yield nightmare
  .evaluate(() => {
    window.localStorage.setItem('testing', 'This will not be persisted');
  })
  .end();

nightmare = Nightmare({
  webPreferences: {
    partition: 'persist: testing'
  }
});
yield nightmare
  .evaluate(() => {
    window.localStorage.setItem('testing', 'This is persisted for other instances with the same paritition name');
  })
  .end();

If you specify a null paritition then it will use the electron default behavior (persistent) or any string that starts with 'persist:' will persist under that partition name, any other string will result in in-memory only storage.

Nightmare is a Node.js module, so you'll need to have Node.js installed. Then you just need to npm install the module:

$ npm install --save nightmare

Nightmare is a node module that can be used in a Node.js script or module. Here's a simple script to open a web page:

import Nightmare from 'nightmare',

const nightmare = Nightmare();

nightmare.goto('http://cnn.com')
  .evaluate(() => {
    return document.title;
  })
  .end()
  .then((title) => {
    console.log(title);
  })

If you save this as cnn.js, you can run it on the command line like this:

npm install --save nightmare
node cnn.js

Nightmare heavily relies on Electron for heavy lifting. And Electron in turn relies on several UI-focused dependencies (eg. libgtk+) which are often missing from server distros.

For help running nightmare on your server distro check out How to run nightmare on Amazon Linux and CentOS guide.

There are three good ways to get more information about what's happening inside the headless browser:

1. Use the DEBUG=* flag described below.
2. Pass { show: true } to the nightmare constructor to have it create a visible, rendered window where you can watch what is happening.
3. Listen for specific events.

To run the same file with debugging output, run it like this DEBUG=nightmare node cnn.js (on Windows use set DEBUG=nightmare & node cnn.js).

This will print out some additional information about what's going on:

nightmare queueing action "goto" +0ms
nightmare queueing action "evaluate" +4ms
Breaking News, U.S., World, Weather, Entertainment & Video News - CNN.com

All nightmare messages

DEBUG=nightmare*

Only actions

DEBUG=nightmare:actions*

Only logs

DEBUG=nightmare:log*

Automated tests for nightmare itself are run using Mocha and Chai, both of which will be installed via npm install. To run nightmare's tests, just run make test.

When the tests are done, you'll see something like this:

make test
  ․․․․․․․․․․․․․․․․․․
  18 passing (1m)

Note that if you are using xvfb, make test will automatically run the tests under an xvfb-run wrapper. If you are planning to run the tests headlessly without running xvfb first, set the HEADLESS environment variable to 0.

WWWWWW||WWWWWW
 W W W||W W W
      ||
    ( OO )__________
     /  |           \
    /o o|    MIT     \
    \___/||_||__||_|| *
         || ||  || ||
        _||_|| _||_||
       (__|__|(__|__|

Copyright (c) 2015 Segment.io, Inc. [email protected]

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the 'Software'), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED 'AS IS', WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.